Computer vision based automatic fruit recognition

The topic of the thesis was analysing the suitability of different algorithms from computer vision for fruit recognition. Fruit provides a challenging domain for recognition due to its diversity of fruits of the same class, the similarity of fruits of different classes and the sheer number of different fruits. For an efficient fruit recognition it was necessary to describe the images with good descriptions. Color, texture, size and shape information of fruits were obtained through a variety of descriptors. The images were then classified based on the obtained descriptors with well known classification methods from machine learning. For a great performance of classification methods a good fruit image database was needed. Since such a publicly available image database does not exist, an image database had to be obtained. Based on the analysis in the thesis a fruit recognition system was built with an 85% success rate on the obtained image database